The MPLS WG Archive

[David Charlap <David.Charlap@marconi.com>]

Bob Anders wrote:
> 
> - MPLS is still used to make forwarding decisions, but the stack
> contains also non-forwarding relevant information either in Exp bits
> or in separate labels (type of information possibly identified by Exp
> bits)

This is done already for certain features.

EXP bits are used to encode priority levels in E-LSPs.

Non-forwarding labels are used to identify VPNs (PWE3) and to identify 
the payload-type for layer-2 traffic (Martini).

> - The MPLS stack is only used as a "vehicle" to encode additional
> non-forwarding relevant information. Forwarding decisions are still
> based on the underlying protocol (e.g. IPv4).

I'm pretty sure that this would violate quite a number of RFCs.  If an 
MPLS router receives labeled traffic, it is supposed to forward that 
traffic according to its label.

If you need to tack additional information onto non-MPLS traffic, there 
are mechanisms for doing so, including options in the IP header and 
various kinds of IP-in-IP encapsulation.

> When talking about non-forwarding relevant data, I am thinking of
> information used for encryption, network statistics, billing, etc ...

WRT MPLS, this information is probably best propagated by the signaling 
stack.  Adding (possibly new) objects to Path and Resv messages works 
well for propagating information between the connection endpoints.  When 
combined with some form of per-LSP traffic counter, you should get all 
the information you need without having to alter the data-plane traffic 
at all.

Of course, your specific application requirements would dictate the best 
approach.

-- David

_______________________________________________
mpls mailing list
mpls@lists.ietf.org
https://www1.ietf.org/mailman/listinfo/mpls